Chapter 14, Part 1, Study Guide
Text: J&M Through 14.3
================================
**F** Marks a possibility for the Final.

Go through the entire lecture notes and study guide, to get a deeper understanding; you will be able to answer the questions on the exam better that way.   Anyway, the purpose of the class is for you to have a good understanding of the material.

Note:  Assignment 4 is about conversion to CNF and applying probabilistic CKY.  Be sure to review the solutions.  However, you won't have to write code for Build_Tree on the final (Question 3 on Assignment 4); you just have to know what it does.  

In a PCFG: 

**F**   Exactly what are the probabilities?  
**F**   Which probabilities must sum to 1?
**F**   How do you calculate the probability of an entire parse tree?
**F**   How do you calculate the probability of a sentence?

Probabilistic CKY parsing:

**F**   Be able to convert a PCFG into Chomsky Normal Form, including  assigning the correct probabilities to the parses
**F**   Be able to apply the probabilistic CKY algorithm including building the tree at the end.


PCFG: Slides 28 and 29 show how you can train a PCFG given tree-banked data.  Do you see exactly what the input and output are, and how the rules and probabilities are derived? 
**F** A question that confirms you know this.

Know that there are methods for training PCFGs without training data, but their performance, to date, are much worse than PCFGs trained on training data.  That's all you need to know about the slide labeled "Inside-Outside".